[HEADSUP] IPX and NWFS to be killed in -current.

[Robert Watson]

On Thu, 9 Dec 2004, Scott Long wrote:

> >>FYI, I have a substantial work in progress in the netperf branch to bring
> >>fine-grained locking to IPX/SPX, as well as to clean up a number of
> >>elements of its implementions (for example, moving the the queue(9) 
> >>macros.  While I'm currently a bit stalled on it due to being overwhelmed
> >>at work (etc), my hope was to get the Giant-free IPX pieces working early
> >>next year.  I think there's a reference to this on the SMPng page showing
> > 
> > 	These are perfect news.  As the former active maintainer of IPX
> > protocol stack and the author of NWFS I'm receive notable amount of complains
> > about IPX support in 5.X as people upgrade boxes from 4.X. For some
> > people it works but nwfs doesn't and vise versa.
> > 
> > 	Addressing phk's request about removal: there was exactly 93
> > questions related to ipx/nwfs in November.  This indeed encourages me to
> > fix them.  Although, I can't promise anything at this point because
> > earning on life doesn't left much free time these days.
> 
> This is wonderful news too!  If you have any questions or need help with
> testing, please don't hesitate to ask. 

The elements I can easily work on here are:

- IPX cleanup and locking, in particular, normalizing the data structures,
  moving to queue(9) from custom lists, locking of the pcb lists and pcbs,
  and interactions with socket locking.  The basic IPX pieces should fall
  out pretty easily; I need to look more closely at SPX before passing
  judgement there.

- Writing some simple IPX/SPX regression suites for loopback traffic and
  ethernet traffic.

I'm less qualified, and insufficiently available, to work with the NCP
protocol pieces, or the file system pieces, and other hands there would be
most welcome.  As with the NFS client, smb client, etc, there's not an
immediate pressure to get netncp and nwfs able to run without the Giant
lock, especially since netncp doesn't use socket upcalls, but as Giant
pushes down on VFS, it's probably an increasingly good idea to do that.  I
anticipate being available to start working on IPX/SPX again sometime in
late January or early Feburary, although I may be able to start on the
regression tests to diagnose any current problems a bit before that.

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert at fledge.watson.org      Principal Research Scientist, McAfee Research